open-mopd
Open-MOPD-Data
Open-MOPD Data
This repository contains the training and evaluation data released with
Open-MOPD, including mixed-domain supervised fine-tuning data, the shared
RL/OPD prompt mixture, and six evaluation benchmarks.
Dataset contents
Configuration
Description
Examples
rl_prompt_mix
Shared math, code, and instruction-following prompts for RL and OPD
86,931
sft_openr1_math_93k
Math SFT data in a unified think-tag format
93,733
sft_ocr_50k
Sampled… See the full description on the dataset page: https://huggingface.co/datasets/BytedTsinghua-SIA/Open-MOPD-Data.marinskyrl-opd-mopd-repro-artifacts
MarinSkyRL OPD/MOPD reproduction artifacts
This public dataset repository preserves the September 2026 MarinSkyRL OPD/MOPD experiment artifacts. artifacts-complete.tar.zst contains the entire original artifacts/ directory, including source checkouts, their Git metadata, the nested local development environment, data, logs, manifests, reports, and symlinks. SHA-256: efcc781203003f67274e23f05e19c6b47679bd761718517ec6241a0e8b0a0b71. The browsable files below are a release copy of… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/marinskyrl-opd-mopd-repro-artifacts.marinskyrl-open-mopd-native-traces
MarinSkyRL native Open-MOPD trace archive
This dataset preserves the retained student trajectories and inline AIME evaluation outputs from the native MarinSkyRL Open-MOPD experiment. The selected step-32 checkpoint is the result of three linked job roots. Files retain the original schema-v3 archive names and directory structure. Source object-store files were copied on 2026-09-18 without rewriting their contents.
Folder
Original run
Coverage
Relation to step-32 policy… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/marinskyrl-open-mopd-native-traces.